The rapid iteration of artificial intelligence (AI) is reshaping firms’ modes of production and the structure of labor-market demand. Meanwhile, the relatively slow adjustment of China’s higher education disciplinary composition has led to a growing structural mismatch between labor supply and industrial demand. Against this background, whether firms’ increasing willingness to develop AI further exacerbates the structural imbalance in employment has emerged as a pressing issue. Using panel data on Chinese A-share-listed firms from 2011 to 2021, this paper systematically examines the impact of firms’ AI development willingness on the structural imbalance between labor supply and demand and investigates the underlying mechanisms.
This paper constructs an indicator of firms’ AI development willingness through textual analysis based on the frequency of AI-related keywords in annual reports. Drawing on the three-sector classification framework, it measures the structural imbalance between labor supply and demand by matching the distribution of university graduates across academic disciplines with industry-level labor demand. To address endogeneity, the number of granted AI-related patents is employed as an instrumental variable, while firms’ actual AI development level, measured by industry-level robot penetration, is introduced to examine the transmission mechanism. The empirical analysis adopts two-way fixed-effects models, supplemented by instrumental-variable estimation, heterogeneity analysis, and robustness tests.
The results show that firms’ AI development willingness significantly exacerbates the structural imbalance between labor supply and demand, and this conclusion remains robust after a series of robustness and endogeneity tests. Further analysis indicates that the imbalance is more pronounced among bachelor’s degree graduates than among associate-degree graduates and postgraduates, and the effect is stronger for non-high-tech firms and heavily polluting firms. Regionally, the effect is significantly positive in eastern and central China but insignificant in western China. Mechanism analysis confirms the transmission path of “development willingness–improvement in actual development level–worsening imbalance,” indicating that firms’ AI development willingness promotes actual AI development, changes labor demand and the skill structure, and thereby aggravates the structural imbalance in employment.
This paper contributes to the literature in four respects. First, it develops an integrated analytical framework jointly considering firm-side labor demand and the supply of talent from higher education to explain the structural imbalance in employment. Second, it constructs a novel firm-level indicator of the mismatch between firms’ labor demand and the disciplinary composition of university graduates. Third, by distinguishing AI development willingness from actual AI development level, it identifies the distinct roles of AI signaling effects and realized technological investments in shaping employment outcomes. Fourth, it provides systematic evidence on heterogeneity across educational attainment, firm type, and region.
The findings suggest that firms should promote responsible AI development, higher education institutions should improve disciplinary adjustment and interdisciplinary talent cultivation, and governments should establish a dynamic matching mechanism between technological change and talent supply to promote the effective alignment among AI development, higher education, and high-quality employment.
利用工具变量对基准模型重新进行回归,结果显示,人工智能发展意愿对就业供求结构失衡的影响依然在1%的水平下显著为正,与基准回归的结果相同。同时,LM统计量为187.156,P<0.01,通过了工具变量不可识别检验;Cragg-Donald Wald F 统计量为188.559,其真实显著性水平不超过10%,通过了弱工具变量检验。因此,该工具变量从理论角度和统计角度上来看都具有良好的性质,并在一定程度上消除了基准回归中可能存在的内生性问题,表明基准回归结论依然成立。
YaoJ Q, ZhangK P, GuoL P, et al. How does artificial intelligence improve firm productivity? Based on the perspective of labor skill structure adjustment[J]. Journal of Management World,2024,40(2):101-116.
[3]
MichaelsG, NatrajA, Van ReenenJ. Has ICT polarized skill demand? Evidence from eleven countries over twenty-five years[J]. The Review of Economics and Statistics,2014,96(1):60-77.
SunX, SongY, ZhaoP Y. The influence of artificial intelligence on the employment of heterogeneous labor force: based on the perspective of labor supply [J]. Inquiry into Economic Issues,2022,43(2):171-190.
HuangX, DongZ Q. How can AI promote high-quality development and employment? [J]. Journal of Central University of Finance & Economics,2023,43(11):3-18.
[8]
FreyC B, OsborneM A. The future of employment: how susceptible are jobs to computerisation?[J]. Technological Forecasting and Social Change,2017,114:254-280.
[9]
ArntzM, GregoryT, ZierahnU. The risk of automation for jobs in OECD countries: a comparative analysis[R]. OECD Social, Employment and Migration Working Papers,No.189,2016.
[10]
GoldinC, KatzL F. The race between education and technology[M].Routledge: Inequality in the 21st Century,2018:380-383.
[11]
AutorD H, DornD. The growth of low-skill service jobs and the polarization of the US labor market[J]. American Economic Review,2013,103(5):1553-1597.
[12]
AlekseevaL, AzarJ, GinéM, et al. The demand for AI skills in the labor market[J]. Labour Economics,2021,71:102002.
[13]
GraetzG, MichaelsG. Robots at work: the impact on productivity and jobs[R]. Centre for Economic Performance, The London School of Economics and Political Science,2015.
[14]
AcemogluD, RestrepoP. Low-skill and high-skill automation[J]. Journal of Human Capital,2018,12(2):204-232.
[15]
LeontiefW. Machines and man[J]. Scientific American,1952,187(3):150-164.
SunZ, HouY L. The influence of artificial intelligence development on industrial total factor productivity: an empirical research based on manufacturing industries in China[J]. Economist,2021,33(1):32-42.
WangY Q, DongW. How the rise of robots has affected China’s labor market: evidence from China’s listed manufacturing firms[J]. Economic Research Journal,2020,55(10):159-175.
KongG W, LiuS S, KongD M. Robots and labor employment: an empirical investigation based on heterogeneity of industries and regions [J]. China Industrial Economics,2020,38(8):80-98.
[22]
GoosM, ManningA L. Lousy and lovely jobs: the rising polarization of work in Britain[J]. The Review of Economics and Statistics,2007,89(1):118-133.
ZhouJ, WangY. Research on the effect of higher education subject supply on industrial economic growth: based on the partial least squares model (PLS)[J]. Hubei Social Sciences,2019,30(6):155-163.
QiuW Q, YueC J. “Useless education” or “the harder, the luckier”: the effect of value-added ability on the labor market entry of undergraduates[J]. Journal of Higher Education,2022,43(10):56-69.
[29]
吴嘉琦,闵维方.教育对产业结构升级的作用机制[J].教育研究,2022,43(1):23-34.
[30]
WuJ Q, MinW F. A probe into the impact of education on industrial structure upgrading[J]. Educational Research,2022,43(1):23-34.
WangP, ChenD X. Artificial intelligence adoption and enterprise resilience: empirical evidence from machine learning to build AI lexicons[J]. Economic Perspectives,2025(9):98-117.
HuangS P. Research on the impact of artificial intelligence technology application on occupational mobility and job satisfaction[J]. Journal of Xi’an Jiaotong University (Social Sciences),2025,45(6):131-140.
[39]
AcemogluD, RestrepoP. Robots and jobs: evidence from US labor markets[J]. Journal of Political Economy,2020,128(6):2188-2244.
FanY G, LiuL, XiaoX R, et al. The relationship between digital capital and labor decision: based on recruitment data [J]. Economic Perspectives,2025,66(9):153-171.
RenS B, HanJ C, LiJ J, et al. Integration, innovation, leadership:serving the construction of the leading country in education [J]. China Higher Education Research,2025,41(7):1-7.